基于多Agent协作的复杂多变生产系统调度研究
文献类型：学位论文
| 作者 | 吕赐兴 |
| 学位类别 | 博士 |
| 答辩日期 | 2007-02-04 |
| 授予单位 | 中国科学院沈阳自动化研究所 |
| 授予地点 | 沈阳自动化研究所 |
| 导师 | 朱云龙 |
| 关键词 | 生产调度 多Agent系统 敏捷调度 多Agent协作 组合拍卖 |
| 其他题名 | Research on Complicated and Changeful Production System Scheduling based on Multi-Agent Coordination |
| 学位专业 | 机械电子工程 |
| 中文摘要 | 近年来，全球化市场的形成及技术变革的加速，制造业面临着新世纪的机遇和挑战，为适应市场竞争，满足生存和发展需要，多种制造模式应运而生。敏捷化智能制造模式是其中的典型代表。该模式强调产品制造过程中企业间在制造执行层的协作，充分利用不同企业的制造资源，达到敏捷响应市场变化和满足用户个性化需求的目的。在敏捷化智能制造模式下，生产系统的管理和控制已超出单个企业的范畴和传统的集中控制方式，更多地体现出复杂、动态和分布协调的特征。这就使得制造系统的管理与控制问题，包括生产调度问题变得更加复杂，无论是管理和控制方式、组织结构，还是决策准则发生了深刻的变革。相应地，要求生产调度系统能够支持生产活动各个环节的合作、资源共享和分配，实现对生产活动合理的组织、调度和管理，对分布自治生产环境下不确定性和突发事件进行处理，并具有快速适应变化和技术更新的能力。面对这些新的特点和要求，如何创建新的适用于敏捷化智能制造模式下的生产系统调度体系，建立起强有力的生产运作机制，将分布于各企业之间的各种制造资源统一起来，协调一致地进行生产，实现共同的生产目标，是生产调度系统所面临的、迫切需要解决的问题。为此，本文采用具有分布协作能力的多Agent技术来构建敏捷化智能制造模式下的生产调度系统，通过Agent的自决策和协作实现生产调度。提出协作上下文的概念以反映生产环境对Agent协作模型和策略的要求，从而实现不同生产环境下Agent的自适应协作。文中进一步提出了具有动态分配能力的基于改进合同网的协作模型和具有规划能力的基于组合拍卖的协作模型，给出了基于拉格朗日松弛的组合拍卖设计。并对给出在系统出现紧急事件的情况下，以切换协作模型和策略的方式来实现的生产系统再调度。具体地，主要的内容如下：本文首先对现有的生产调度研究成果进行了回顾和总结，指出在新的制造形势下生产调度系统体现出复杂、动态和分布协调的特征。面对这些特征和要求，不仅需要研究新的、更实用的调度算法，还需要从系统的角度出发，研究适应复杂动态生产系统调度需求的系统模型、调度机制以及相应的调度策略和方法。在已有研究中，基于多Agent的生产调度方法研究已经取得了一定的成果，但存在协作模型和策略与生产调度问题存在紧密的耦合、难以满足复杂多变的生产调度要求的问题。因此，充分利用和发挥多Agent技术的协作优势，以不同的协作模型和策略适应不同的生产场景，从而提高Agent协作的适应能力，成为研究基于Agent的生产调度中的一个重大而又迫切的课题，这也是本文研究工作的出发点和目标所在。本文分别从生产系统到多Agent系统的映射、Agent的结构模型、多Agent系统的体系结构和面向生产调度的多Agent协作机制和策略等几个方面对Agent技术在生产调度中的应用的研究现状、存在的问题和发展趋势进行了简单回顾和分析。在此基础上，给出了本文采用的基于Agent的生产系统建模方法，该方法兼顾产品制造过程和生产资源管理，采用混合建模方法，既包含资源Agent和任务Agent，又包括功能Agent和管理Agent。并给出了Agent的定义和功能，完成了各类Agent的结构搭建和描述。组织控制结构在制造系统中起着举足轻重的作用。本文对制造系统控制结构的研究现状和发展趋势进行了深入研究，提出了一种分布式结构与递阶结构相集成的混合控制模型与结构。在该结构中，通过引入调解Agent方法减少系统中Agent间的协调活动开销，提供了开放的、可扩展的框架结构，确保了系统的稳定性和可扩展性。完成系统组织结构和Agent模型的构建之后，如何有效地控制和协调系统内各个Agent的行为，获得满意的调度结果是一个非常棘手却又至关重要的问题。为此，完成了从生产调度问题到多Agent协作问题的映射，将生产调度问题映射成了一个多Agent协作问题。同时，提出了协作上下文的概念，并建立了面向生产调度的多Agent协作框架。在该框架的支持下，Agent根据不同的协作上下文采用不同的协作模型和策略进行协作，最终完成生产调度目标。混合式调度系统控制结构和基于Agent协作的调度框架，为解决生产调度问题建立了良好的支撑环境，然而，系统目标的最终实现，还有赖于良好的协作模型和策略。因此，对具有动态反应能力的基于改进合同网的协作模型以及具有一定规划能力的基于组合拍卖的协作模型进行了研究。以具有路径柔性、in-tree次序约束的生产调度问题为研究对象，给出了基于拉格朗日松弛的组合拍卖设计。并以系统的扰动处理为例，探讨了系统如何通过选择不同的协作模型以适应不同的协作上下文，给出了系统切换协作模型的详细过程。此外，在对Agent模型、系统组织结构、协作框架和协作模型研究的基础上，利用JADE平台，设计并开发了一个基于Agent的生产调度原型系统，并将前面的研究成果应用到该原型系统中。在实现多Agent应用系统的模块化、提高软件原型的可操作性和易用性的基础上，通过对基于Agent的生产调度仿真实验，验证了基于多Agent协作生产调度的可行性和有效性，以及系统在车间扰动下的动态调度能力。文章最后对已取得的研究成果进行了总结，并对进一步需要研究的问题进行了展望。 |
| 索取号 | TP277/L93/2007 |
| 英文摘要 | Production scheduling in manufacturing has been a classical problem hotly researched in manufacturing field. At the origin researchers’ attention was focused on scheduling optimality. After that, however, it shifted to scheduling flexibility, and more recently to the support for distribution and coordination. This is the results of challenges provoked by requirements for global competitiveness in open markets, and continuous development of both manufacturing technologies and socio-organizational criteria. Many new capabilities are needed in distributed and coordinated scheduling systems, such as the capability to react dynamically when events can not be foreseen in the current schedule. As a result, new methods and mechanisms are expected for the agile production-scheduling problem. In this paper, a multi-agent coordination based scheduling system is presented to solve the agile scheduling problem, and some good results have been obtained. After the survey and analysis on production scheduling research, distribution and coordination features of production scheduling are addressed. Because of the complexity of distributed and coordinated manufacturing, new algorithms alone can not deal with this problem. Proper production scheduling models, scheduling mechanisms, corresponding policies and methods are all needed. As a kind of approach for designing distributed and coordinated production scheduling, agent-based production scheduling was successful to some extent in the past researches. However, agent technology and its application are under development. There have been few successful cases up to now. Furthermore, agent-based production systems are currently designed just aiming at solving special problems in production scheduling. It is difficult for them to deal with the complexity of production scheduling. Therefore, how to take advantages of multi-agent systems, how to select coordination models and coordination policies according to production context, how to improve the adaptability and coordination of agents, these are all intensively considered in this paper. As far as agent-based production scheduling is concerned, previous relative work by others are surveyed at first, which including mapping from production system to multi-agent system, structure model of agents, architecture of multi-agent system, coordination mechanism and policy etc. Their defects and development trends in the near future are also pointed out. Based on them, a hybrid modeling method for agent-based production system is present. In this method, both manufacturing resource and production process are considered, and all kinds of agents, including their functions, are elaborately defined. There are resource agents, task agents, function agents and management agents in the system. System architecture plays a very important role in the agent-basesd manufacturing system. An architecture integrating distributed architecture and hierarchy architecture is present after a survey on the past works. In the architecture, a kind of agents, called mediator agent, is introduced into system. It can distinctly decrease the cost of coordination and make the framework open and extendible. The hybrid architecture and scheduling framework have constituted supporting environments for solving production scheduling problem. However, the achievement of system goals also tightly relies on coordination model and coordination policy. Therefore, two coordination models are presented in the paper. One is on the basis of modified contract net, and the other on combinatorial auction. The former has the capability of reacting to dynamic events, and the latter has the capability of planning and optimization in a time window. Under the circumstances of existing disturbance, the process of how system switch different coordination models in different coordination context is analyzed. For examination, a prototype system of agent-based production scheduling is designed, which is developed on JADE platform. This system mainly focuses on scheduling problems in which tasks have alternative routes and in-tree precedence constraint. Calculation test shows that the system performs very well with high flexibility and efficiency. When added disturbances such as resource break-down, emergency task and task cancelled, the capability of handling dynamic environments is displayed. In the conclusions of this thesis, research results are introduced. Future research works are forecast and simply described. |
| 语种 | 中文 |
| 公开日期 | 2010-11-29 |
| 产权排序 | 1 |
| 页码 | 118 |
| 分类号 | TP277 |
| 源URL | [http://ir.sia.ac.cn//handle/173321/79]  |
| 专题 | 沈阳自动化研究所_工业信息学研究室_先进制造技术研究室 |
| 推荐引用方式 GB/T 7714 | 吕赐兴. 基于多Agent协作的复杂多变生产系统调度研究[D]. 沈阳自动化研究所. 中国科学院沈阳自动化研究所. 2007. |
入库方式： OAI收割
来源：沈阳自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。